Correct length in sharding

[shuningjin]

Description

Due to change in #2023, "activation_length" definition changed. Specifically, "expert" is added there, and causing some conflicts:

• As a result, it should not be used together with "activation_batch", otherwise sharding like CP/SP will be ignored. See details in b/441547754.
• Similarly, should not be used together with "activation_embed_and_logits_batch". e.g, b/433561718#comment22

1. For conflicting combination "activation_batch" + "activation_length" in file. Either update to "activation_length_no_exp" or "activation_norm_length".

• we can restore old definition by using "activation_length_no_exp"
• In some cases, they should be "activation_norm_length", aligning with Tensor Sequence Parallelism (TSP).

2. For conflicting combination "activation_embed_and_logits_batch", "activation_length", we can restore old definition by using "activation_length_no_exp"

FIXES: b/441547754

1 Changes related to "activation_batch"

TSP should use "activation_norm_length" for decoder block

• as in TSP PR: introduce activation_norm_length, see changes in 1, 3, 5
• other decoder blocks also uses this practice, e.g., deepseek, etc
• qwen3.py, decoders.py

TSP should use "activation_length_no_exp" for MLP and MoE

• as in TSP PR: remain as activation_length, which now should be activation_length_no_exp
• linears.py: restore, activation_length -> activation_length_no_exp
• moe.py: restore, activation_length -> PR2023 -> activation_norm_length -> activation_length_no_exp

Additional changes

• gpt3.py: attention layer, restore
• test: restore
• pyconfig.py: correct typo

2 Changes related to "activation_embed_and_logits_batch"

train.py, decoder.py, embedding.py: restore to "activation_length_no_exp"

Tests

N/A

Checklist

Before submitting this PR, please make sure (put X in square brackets):

• I have performed a self-review of my code.
• I have necessary comments in my code, particularly in hard-to-understand areas.
• I have run end-to-end tests tests and provided workload links above if applicable.
• I have made or will make corresponding changes to the doc if needed.

[github-actions]

This PR has been automatically marked as stale because it has not had recent activity. It will be closed soon if no further activity occurs. Thank you for your contributions.

[github-actions]

This PR was closed because it has been inactive for a while. Please reopen it if you are still working on it.